Would you like to inspect the original subtitles? These are the user uploaded subtitles that are being translated:
1
00:00:00,890 --> 00:00:03,389
this is the screencast tutorial for how
2
00:00:03,389 --> 00:00:05,520
to use the program g-power to determine
3
00:00:05,520 --> 00:00:07,830
two things first of all is a priori
4
00:00:07,830 --> 00:00:09,690
power that is computing the required
5
00:00:09,690 --> 00:00:12,000
sample size that you need to achieve a
6
00:00:12,000 --> 00:00:14,639
desired level of power for a study given
7
00:00:14,639 --> 00:00:17,369
levels of alpha and effect size and the
8
00:00:17,369 --> 00:00:19,170
other thing is to compute post hoc power
9
00:00:19,170 --> 00:00:21,480
that is exactly what is the obtained
10
00:00:21,480 --> 00:00:22,890
level of power you have in the study
11
00:00:22,890 --> 00:00:25,710
given your sample statistics to estimate
12
00:00:25,710 --> 00:00:27,660
things like effect size as well as the
13
00:00:27,660 --> 00:00:29,240
Alpha level that you use in your study
14
00:00:29,240 --> 00:00:32,159
now the nice thing about G power is as
15
00:00:32,159 --> 00:00:34,410
you can see here there are a variety of
16
00:00:34,410 --> 00:00:36,239
statistical tests that you can select
17
00:00:36,239 --> 00:00:39,270
from so we're not going to cover all of
18
00:00:39,270 --> 00:00:41,430
these this semester or last semester or
19
00:00:41,430 --> 00:00:42,780
even over the course of this curriculum
20
00:00:42,780 --> 00:00:44,579
okay but there really is a lot of
21
00:00:44,579 --> 00:00:46,890
flexibility in the program G power for
22
00:00:46,890 --> 00:00:48,440
any design that you might come up with
23
00:00:48,440 --> 00:00:50,910
there's different test families as well
24
00:00:50,910 --> 00:00:52,110
right now we're going to stick with just
25
00:00:52,110 --> 00:00:55,079
the T tests but as you can see F tests
26
00:00:55,079 --> 00:00:57,059
which may be required for your upcoming
27
00:00:57,059 --> 00:00:59,640
project are available in here as well in
28
00:00:59,640 --> 00:01:02,370
addition to things like chi-square and D
29
00:01:02,370 --> 00:01:04,920
tests that we've done in the past again
30
00:01:04,920 --> 00:01:06,030
what we're going to do to stick with T
31
00:01:06,030 --> 00:01:07,860
tests for now and let's look at the
32
00:01:07,860 --> 00:01:09,990
first example say that what we want to
33
00:01:09,990 --> 00:01:12,180
do is to determine what sample size will
34
00:01:12,180 --> 00:01:14,250
be required to achieve a desired level
35
00:01:14,250 --> 00:01:18,540
of power for a paired samples t-test so
36
00:01:18,540 --> 00:01:20,400
we know we're in the t-test family and
37
00:01:20,400 --> 00:01:22,650
then we go over to statistical tests and
38
00:01:22,650 --> 00:01:24,990
find the test in which we're interested
39
00:01:24,990 --> 00:01:27,600
in this case a paired samples are what
40
00:01:27,600 --> 00:01:29,400
they're calling a matched pairs t-test
41
00:01:29,400 --> 00:01:31,710
is indeed what we're looking at the
42
00:01:31,710 --> 00:01:33,479
difference between two dependent means
43
00:01:33,479 --> 00:01:36,360
so we select that test now as I
44
00:01:36,360 --> 00:01:38,820
mentioned in G power all you have to do
45
00:01:38,820 --> 00:01:40,619
is to fill in the different values that
46
00:01:40,619 --> 00:01:41,970
you know and it will calculate the
47
00:01:41,970 --> 00:01:44,820
unknown value for you so in this case
48
00:01:44,820 --> 00:01:46,560
the first thing we want to do is to tell
49
00:01:46,560 --> 00:01:48,360
it what type of tests that we're looking
50
00:01:48,360 --> 00:01:54,540
at a priori or pose talk now you can see
51
00:01:54,540 --> 00:01:55,890
there are a lot of other options in here
52
00:01:55,890 --> 00:01:57,270
as well but these are going to be the
53
00:01:57,270 --> 00:01:59,040
only two in which we're interested for
54
00:01:59,040 --> 00:02:00,750
now let's do the a priori test which is
55
00:02:00,750 --> 00:02:02,369
going to tell us what is our required
56
00:02:02,369 --> 00:02:04,590
sample size to achieve a desired level
57
00:02:04,590 --> 00:02:07,500
of power then we need to do is to let it
58
00:02:07,500 --> 00:02:09,000
know whether it's a directional or a non
59
00:02:09,000 --> 00:02:10,949
directional hypothesis by telling it the
60
00:02:10,949 --> 00:02:12,870
number of tails we know that a non
61
00:02:12,870 --> 00:02:13,710
directional test
62
00:02:13,710 --> 00:02:16,320
a two-tailed test a directional test is
63
00:02:16,320 --> 00:02:18,660
a one-tailed test let's assume for now
64
00:02:18,660 --> 00:02:19,530
that we're going to going with the
65
00:02:19,530 --> 00:02:22,830
two-tailed test now as mentioned in
66
00:02:22,830 --> 00:02:25,410
lecture all we then to know are what are
67
00:02:25,410 --> 00:02:28,440
our effect size our alpha level and our
68
00:02:28,440 --> 00:02:31,230
desired power and then what we can do is
69
00:02:31,230 --> 00:02:33,360
to then compute the sample size that
70
00:02:33,360 --> 00:02:34,920
would be necessary in order to achieve
71
00:02:34,920 --> 00:02:38,160
that level of power so the effect size
72
00:02:38,160 --> 00:02:42,630
here at 0.5 is a moderate effect size we
73
00:02:42,630 --> 00:02:44,160
can simply change that by 2 anything we
74
00:02:44,160 --> 00:02:45,300
want if we think it'll be a little bit
75
00:02:45,300 --> 00:02:48,180
larger 0.6 if we think it would be a
76
00:02:48,180 --> 00:02:52,140
small effect size of 0.2 0.8 for a large
77
00:02:52,140 --> 00:02:53,760
effect size those are conventional
78
00:02:53,760 --> 00:02:55,500
levels so let's just stick with a
79
00:02:55,500 --> 00:02:58,380
moderate effect size of 0.5 our alpha
80
00:02:58,380 --> 00:03:00,330
level is indeed going to be a value of
81
00:03:00,330 --> 00:03:03,120
0.05 as it is always going to be in this
82
00:03:03,120 --> 00:03:04,830
course and in behavioral science in
83
00:03:04,830 --> 00:03:07,020
general and then what we can do is to
84
00:03:07,020 --> 00:03:08,370
say what is the level of power that we
85
00:03:08,370 --> 00:03:10,530
would desire now point nine five is a
86
00:03:10,530 --> 00:03:12,120
great level of power recall what this
87
00:03:12,120 --> 00:03:14,160
means is that if there really is an
88
00:03:14,160 --> 00:03:16,680
effect we want there to be a 95% chance
89
00:03:16,680 --> 00:03:19,520
of detecting this effect in our study
90
00:03:19,520 --> 00:03:21,810
now we might not want such a strange
91
00:03:21,810 --> 00:03:23,190
requirement let's lower this down to
92
00:03:23,190 --> 00:03:26,160
something like 90% and that's all it
93
00:03:26,160 --> 00:03:27,750
takes is inputting your effect size
94
00:03:27,750 --> 00:03:31,830
estimate D your alpha level and your
95
00:03:31,830 --> 00:03:34,500
desired power and then clicking on
96
00:03:34,500 --> 00:03:37,680
calculate now what it's going to produce
97
00:03:37,680 --> 00:03:39,990
is a graph similar to the one that we
98
00:03:39,990 --> 00:03:41,970
saw in lecture as well ok where it's
99
00:03:41,970 --> 00:03:43,610
showing you the red distribution
100
00:03:43,610 --> 00:03:46,200
represents the null distribution the
101
00:03:46,200 --> 00:03:48,600
dashed blue distribution represents the
102
00:03:48,600 --> 00:03:49,950
potential alternative research
103
00:03:49,950 --> 00:03:52,170
hypothesis and you can see the shaded
104
00:03:52,170 --> 00:03:54,630
areas for alpha and beta similar to the
105
00:03:54,630 --> 00:03:56,400
way they were discussed in lecture now
106
00:03:56,400 --> 00:03:57,810
what's going to be important for us then
107
00:03:57,810 --> 00:03:59,340
is looking at what is going to be the
108
00:03:59,340 --> 00:04:00,900
sample size that's required to achieve
109
00:04:00,900 --> 00:04:03,420
this level of power the place you can
110
00:04:03,420 --> 00:04:05,940
find that is right here under total
111
00:04:05,940 --> 00:04:08,430
sample size and we can see that we need
112
00:04:08,430 --> 00:04:10,550
a total of 44 participants in this study
113
00:04:10,550 --> 00:04:13,800
that is in a within subject situation
114
00:04:13,800 --> 00:04:15,510
where we're comparing two dependent
115
00:04:15,510 --> 00:04:17,850
means we would need forty four people
116
00:04:17,850 --> 00:04:20,130
and then of course we know this would be
117
00:04:20,130 --> 00:04:21,899
within subject study where these 44
118
00:04:21,899 --> 00:04:23,040
people are participating in both
119
00:04:23,040 --> 00:04:26,430
conditions now there's a couple other
120
00:04:26,430 --> 00:04:27,580
things we want to look at
121
00:04:27,580 --> 00:04:29,409
here in particular one thing we can do
122
00:04:29,409 --> 00:04:31,750
is to look at the XY plot for a range of
123
00:04:31,750 --> 00:04:35,909
possible values if we want to do that
124
00:04:35,909 --> 00:04:38,319
then all you need to do is to click on
125
00:04:38,319 --> 00:04:39,969
the button at the bottom that brings up
126
00:04:39,969 --> 00:04:42,250
this plotting window and you can change
127
00:04:42,250 --> 00:04:44,080
everything you want in here as well you
128
00:04:44,080 --> 00:04:46,810
can change the effect size you can
129
00:04:46,810 --> 00:04:49,270
change your alpha level and you can set
130
00:04:49,270 --> 00:04:51,669
some other specific details as well for
131
00:04:51,669 --> 00:04:53,800
example say that what you want to see is
132
00:04:53,800 --> 00:04:55,270
the sample size required to achieve
133
00:04:55,270 --> 00:04:57,370
different levels of power ranging from
134
00:04:57,370 --> 00:05:04,960
0.5 all the way up to 0.95 once you set
135
00:05:04,960 --> 00:05:06,789
the details in the values how you like
136
00:05:06,789 --> 00:05:07,930
them then all you have to do is click
137
00:05:07,930 --> 00:05:09,669
draw plot and then it will show you the
138
00:05:09,669 --> 00:05:13,659
values specifically to achieve a certain
139
00:05:13,659 --> 00:05:15,310
level of power which is shown on the
140
00:05:15,310 --> 00:05:18,340
x-axis say that we want to achieve power
141
00:05:18,340 --> 00:05:21,759
of 0.8 we can use this plot then to see
142
00:05:21,759 --> 00:05:23,440
how many participants are going to be
143
00:05:23,440 --> 00:05:26,229
required if we read over on the y-axis
144
00:05:26,229 --> 00:05:28,270
it's going to show us about 33
145
00:05:28,270 --> 00:05:30,460
participants are required total sample
146
00:05:30,460 --> 00:05:34,509
size given our effect size and alpha
147
00:05:34,509 --> 00:05:37,810
level in order to achieve that power of
148
00:05:37,810 --> 00:05:44,199
0.8 in a trailer that simple there's a
149
00:05:44,199 --> 00:05:46,810
second example we can look at as well so
150
00:05:46,810 --> 00:05:48,370
say that you don't have your effect size
151
00:05:48,370 --> 00:05:50,680
in particular for example what you may
152
00:05:50,680 --> 00:05:52,599
be doing is basing your effect size off
153
00:05:52,599 --> 00:05:54,969
of previous research so you run one
154
00:05:54,969 --> 00:05:56,650
study things didn't go so well
155
00:05:56,650 --> 00:05:59,349
yet you have the data the sample data
156
00:05:59,349 --> 00:06:00,759
from that previous study that you can
157
00:06:00,759 --> 00:06:02,349
use to estimate effect size for an
158
00:06:02,349 --> 00:06:05,650
upcoming study for example any situation
159
00:06:05,650 --> 00:06:06,669
like this where you want to use the
160
00:06:06,669 --> 00:06:08,289
sample statistics to determine your
161
00:06:08,289 --> 00:06:10,300
effect size can be done by clicking on
162
00:06:10,300 --> 00:06:11,940
this determined button here on the left
163
00:06:11,940 --> 00:06:13,960
what that's going to do is to open up
164
00:06:13,960 --> 00:06:16,690
this little side window or drawer and in
165
00:06:16,690 --> 00:06:18,639
here what you can do is to calculate the
166
00:06:18,639 --> 00:06:21,339
sample statistics enter them in directly
167
00:06:21,339 --> 00:06:23,349
and then it will determine the effect
168
00:06:23,349 --> 00:06:26,919
size for you okay so for example we know
169
00:06:26,919 --> 00:06:28,330
that when we're calculating a paired
170
00:06:28,330 --> 00:06:30,400
samples t-test we end up calculating a
171
00:06:30,400 --> 00:06:33,969
mean different score and a standard
172
00:06:33,969 --> 00:06:36,459
deviation of that different score so if
173
00:06:36,459 --> 00:06:37,539
you come and select this radio button
174
00:06:37,539 --> 00:06:40,479
for from differences and say that we've
175
00:06:40,479 --> 00:06:41,440
done a study where we find
176
00:06:41,440 --> 00:06:44,920
mean difference of say two and a
177
00:06:44,920 --> 00:06:46,090
standard deviation among different
178
00:06:46,090 --> 00:06:49,870
scores of say five point five then we
179
00:06:49,870 --> 00:06:51,280
can click this calculate button here at
180
00:06:51,280 --> 00:06:52,900
the bottom and it will do the effect
181
00:06:52,900 --> 00:06:55,030
size calculation for us now this is a
182
00:06:55,030 --> 00:06:56,530
pretty simple calculation which I'm sure
183
00:06:56,530 --> 00:06:58,270
you can verify using a calculator as
184
00:06:58,270 --> 00:07:00,460
well but if you have these sample
185
00:07:00,460 --> 00:07:02,200
statistics handy this is a convenient
186
00:07:02,200 --> 00:07:03,940
way for it to calculate the effect size
187
00:07:03,940 --> 00:07:06,100
and then by clicking the other button
188
00:07:06,100 --> 00:07:08,710
transfer to main window then you can see
189
00:07:08,710 --> 00:07:10,090
what it does is it copies the effect
190
00:07:10,090 --> 00:07:11,890
size into the main window of g-power
191
00:07:11,890 --> 00:07:14,650
directly for you from there everything
192
00:07:14,650 --> 00:07:16,600
else perceived as normal you can simply
193
00:07:16,600 --> 00:07:19,510
click calculate to determine your total
194
00:07:19,510 --> 00:07:21,250
sample size necessary in this case it
195
00:07:21,250 --> 00:07:23,620
would be 82 you can see as the effect
196
00:07:23,620 --> 00:07:26,860
size went down from 0.5 to 0.36 the
197
00:07:26,860 --> 00:07:29,320
sample size required went up to maintain
198
00:07:29,320 --> 00:07:32,470
the same level of power and once again
199
00:07:32,470 --> 00:07:34,510
you can click for the XY plot to show an
200
00:07:34,510 --> 00:07:43,510
entire range of values so let's look at
201
00:07:43,510 --> 00:07:45,220
one other type of test the independent
202
00:07:45,220 --> 00:07:47,140
samples t-test because although the
203
00:07:47,140 --> 00:07:48,640
logic is the same and I'm sure you could
204
00:07:48,640 --> 00:07:49,870
figure it out for yourself but it is
205
00:07:49,870 --> 00:07:51,580
handled a little bit differently so
206
00:07:51,580 --> 00:07:52,750
let's take a look at that one now as
207
00:07:52,750 --> 00:07:55,270
well again this is still going to be a
208
00:07:55,270 --> 00:07:57,790
t-test but specifically this is going to
209
00:07:57,790 --> 00:07:59,740
be differences between means of two
210
00:07:59,740 --> 00:08:03,190
independent groups this then is the
211
00:08:03,190 --> 00:08:07,810
independent samples t-test in g-power so
212
00:08:07,810 --> 00:08:10,120
clicking on here again what we may still
213
00:08:10,120 --> 00:08:11,800
want to know is what is the sample size
214
00:08:11,800 --> 00:08:13,990
that's going to be required to achieve a
215
00:08:13,990 --> 00:08:17,470
specific desired level of power so again
216
00:08:17,470 --> 00:08:19,030
everything is the same so let's look at
217
00:08:19,030 --> 00:08:21,880
the same effect size of 0.5 alpha of
218
00:08:21,880 --> 00:08:25,600
0.05 power again and let's say 0.9 0
219
00:08:25,600 --> 00:08:27,580
just changing the values in the boxes
220
00:08:27,580 --> 00:08:30,190
here okay the only other box that's on
221
00:08:30,190 --> 00:08:32,760
here is the allocation ratio n2 to n1
222
00:08:32,760 --> 00:08:35,110
now specifically what we're going to
223
00:08:35,110 --> 00:08:36,789
strive for in an independent samples
224
00:08:36,789 --> 00:08:39,309
test is to have equal sample sizes in
225
00:08:39,309 --> 00:08:41,349
the two groups well if that's the case
226
00:08:41,349 --> 00:08:43,210
in the ratio of sample size between the
227
00:08:43,210 --> 00:08:45,670
groups is going to be equal to one but
228
00:08:45,670 --> 00:08:47,350
it may not always be the case for now
229
00:08:47,350 --> 00:08:49,210
let's assume it is and if it is you can
230
00:08:49,210 --> 00:08:51,310
simply click calculate and then it's
231
00:08:51,310 --> 00:08:52,660
going to show you the sample size in
232
00:08:52,660 --> 00:08:54,940
each group 70 instead
233
00:08:54,940 --> 00:08:59,140
for a total sample size of 140 now if
234
00:08:59,140 --> 00:09:00,520
you have a different allocation ratio
235
00:09:00,520 --> 00:09:03,010
let's say that you have different
236
00:09:03,010 --> 00:09:04,570
numbers of people within the two
237
00:09:04,570 --> 00:09:07,060
different groups say that it's an
238
00:09:07,060 --> 00:09:10,600
allocation issue of 0.75 okay this would
239
00:09:10,600 --> 00:09:12,310
be an example if you have say 30 people
240
00:09:12,310 --> 00:09:14,500
in one group and 40 people in the other
241
00:09:14,500 --> 00:09:17,440
group well then the number of people in
242
00:09:17,440 --> 00:09:20,020
one group 30 over the number of people
243
00:09:20,020 --> 00:09:23,380
in the other group 40 would equal 3/4 or
244
00:09:23,380 --> 00:09:26,740
0.75 in this case you still just click
245
00:09:26,740 --> 00:09:29,110
calculate and again it's going to show
246
00:09:29,110 --> 00:09:31,150
you the sample size that you need which
247
00:09:31,150 --> 00:09:33,190
you can see is student 42 here and then
248
00:09:33,190 --> 00:09:35,020
how they're going to be allocated across
249
00:09:35,020 --> 00:09:37,240
the two different groups now you may
250
00:09:37,240 --> 00:09:38,710
notice that the total sample size
251
00:09:38,710 --> 00:09:40,510
necessary in this case has gone up a
252
00:09:40,510 --> 00:09:43,330
little bit okay this is an important
253
00:09:43,330 --> 00:09:44,860
point in an independent samples t-test
254
00:09:44,860 --> 00:09:46,750
that you actually achieve the highest
255
00:09:46,750 --> 00:09:49,840
level of power if your sample your total
256
00:09:49,840 --> 00:09:51,850
sample is allocated evenly across the
257
00:09:51,850 --> 00:09:57,040
two groups now another thing that we can
258
00:09:57,040 --> 00:10:01,330
do here is to think about using the
259
00:10:01,330 --> 00:10:04,150
determined window in order to enter the
260
00:10:04,150 --> 00:10:05,950
sample statistics just like we did for
261
00:10:05,950 --> 00:10:09,340
the paired samples test okay now in this
262
00:10:09,340 --> 00:10:11,080
case again what we may want to do is
263
00:10:11,080 --> 00:10:12,670
just enter the sample data say we've
264
00:10:12,670 --> 00:10:14,170
collected some data we find the mean of
265
00:10:14,170 --> 00:10:17,230
the first group is 55 mean of the second
266
00:10:17,230 --> 00:10:21,580
group is 45 Center deviation 1 could be
267
00:10:21,580 --> 00:10:24,280
6 say the other group is 7 we can use
268
00:10:24,280 --> 00:10:25,840
this to calculate our effect size here
269
00:10:25,840 --> 00:10:30,210
as well transfer it to the main window
270
00:10:30,210 --> 00:10:33,400
which it's done and then complete the
271
00:10:33,400 --> 00:10:37,510
calculation for us okay in this case of
272
00:10:37,510 --> 00:10:39,130
course with such a large effect size we
273
00:10:39,130 --> 00:10:40,660
see that we need a relatively fewer
274
00:10:40,660 --> 00:10:46,000
people in the two groups now if we don't
275
00:10:46,000 --> 00:10:48,430
have equal effect sizes which is the top
276
00:10:48,430 --> 00:10:50,230
radio button here in this calculation or
277
00:10:50,230 --> 00:10:52,780
determine drawer then what we may do is
278
00:10:52,780 --> 00:10:55,120
to enter in the mean of the two groups
279
00:10:55,120 --> 00:10:58,630
again let's say it's 55 and 45 then what
280
00:10:58,630 --> 00:10:59,890
it asks for here is the standard
281
00:10:59,890 --> 00:11:02,350
deviation within each group now it's
282
00:11:02,350 --> 00:11:03,820
important to know what this is going to
283
00:11:03,820 --> 00:11:05,950
be is essentially the pooled standard
284
00:11:05,950 --> 00:11:08,390
deviation that is the school
285
00:11:08,390 --> 00:11:11,570
a root of the pooled variance estimate
286
00:11:11,570 --> 00:11:15,380
that we know how to calculate let's say
287
00:11:15,380 --> 00:11:17,750
that's something like six point two just
288
00:11:17,750 --> 00:11:19,670
to select a number again we can
289
00:11:19,670 --> 00:11:22,640
calculate the effect size transfer it to
290
00:11:22,640 --> 00:11:26,180
the main window and then calculate the
291
00:11:26,180 --> 00:11:28,700
sample size necessary to produce the
292
00:11:28,700 --> 00:11:31,160
desired level of power once again with
293
00:11:31,160 --> 00:11:33,680
such a large effect size the sample size
294
00:11:33,680 --> 00:11:35,030
of a needing each group is relatively
295
00:11:35,030 --> 00:11:38,660
pretty small so finally let's look at an
296
00:11:38,660 --> 00:11:40,670
example of how we can then calculate
297
00:11:40,670 --> 00:11:45,260
post-talk power now remember this is
298
00:11:45,260 --> 00:11:46,250
going to change a little bit because
299
00:11:46,250 --> 00:11:48,470
what we're looking at finding now is not
300
00:11:48,470 --> 00:11:50,150
for a given level of power what is our
301
00:11:50,150 --> 00:11:53,900
sample size but given our sample size
302
00:11:53,900 --> 00:11:57,080
what is our achieved level of power now
303
00:11:57,080 --> 00:11:58,280
for this we're just going to stick with
304
00:11:58,280 --> 00:12:00,980
independent samples t-test and doing it
305
00:12:00,980 --> 00:12:02,480
for the dependent test is very similar
306
00:12:02,480 --> 00:12:03,770
so I'm not going to go back through the
307
00:12:03,770 --> 00:12:05,810
motions there but again where you can
308
00:12:05,810 --> 00:12:08,120
see it has happened to retain the values
309
00:12:08,120 --> 00:12:09,860
that we calculated in our last analysis
310
00:12:09,860 --> 00:12:12,770
this may or may not be the case okay but
311
00:12:12,770 --> 00:12:14,210
what you can do in this case especially
312
00:12:14,210 --> 00:12:16,010
is because you're doing this post hoc
313
00:12:16,010 --> 00:12:18,080
you're typically going to have exactly
314
00:12:18,080 --> 00:12:19,610
the values that would go in these boxes
315
00:12:19,610 --> 00:12:22,370
over here okay so just to change the
316
00:12:22,370 --> 00:12:23,570
numbers to produce a different example
317
00:12:23,570 --> 00:12:26,330
say we have one class that get 275
318
00:12:26,330 --> 00:12:29,480
average on an exam another class get 272
319
00:12:29,480 --> 00:12:31,670
average we have the standard deviation
320
00:12:31,670 --> 00:12:38,870
for the exam scores in each group if
321
00:12:38,870 --> 00:12:39,950
that's the case again we can just
322
00:12:39,950 --> 00:12:41,750
calculate the effect size so these are
323
00:12:41,750 --> 00:12:42,740
going to be values that are coming
324
00:12:42,740 --> 00:12:44,270
directly out of your analysis either
325
00:12:44,270 --> 00:12:46,190
through SPSS or Excel or wherever you're
326
00:12:46,190 --> 00:12:48,650
going to calculate them once again all
327
00:12:48,650 --> 00:12:49,760
we have to do is transfer to the main
328
00:12:49,760 --> 00:12:53,480
window okay and now then we're also
329
00:12:53,480 --> 00:12:54,920
going to know the sample we had in each
330
00:12:54,920 --> 00:12:56,750
group let's say that we had 18 people in
331
00:12:56,750 --> 00:13:01,250
the first group and 20 people in the
332
00:13:01,250 --> 00:13:04,310
second group now in this case when we
333
00:13:04,310 --> 00:13:06,020
calculate it's going to show us is the
334
00:13:06,020 --> 00:13:08,300
calculated the achieved level of power
335
00:13:08,300 --> 00:13:11,710
in this case rather low it's 0.27 and
336
00:13:11,710 --> 00:13:13,790
you can see that in terms of how we
337
00:13:13,790 --> 00:13:15,380
talked about it in lecture especially
338
00:13:15,380 --> 00:13:18,980
with the overlap here of the alternative
339
00:13:18,980 --> 00:13:20,660
distribution and specifically how much
340
00:13:20,660 --> 00:13:21,680
that is
341
00:13:21,680 --> 00:13:23,630
the left of our critical value which is
342
00:13:23,630 --> 00:13:27,860
shown in green well that's the last
343
00:13:27,860 --> 00:13:29,420
example that I wanted to go through what
344
00:13:29,420 --> 00:13:30,800
we've seen now is how to do two things
345
00:13:30,800 --> 00:13:33,680
how to calculate a priority power that
346
00:13:33,680 --> 00:13:35,600
is for a given desired level of power
347
00:13:35,600 --> 00:13:38,060
what is the sample size necessary to
348
00:13:38,060 --> 00:13:41,630
achieve it and post-hoc power given all
349
00:13:41,630 --> 00:13:43,459
of our sample statistics that is the
350
00:13:43,459 --> 00:13:44,839
values that are associated with our
351
00:13:44,839 --> 00:13:47,180
exact experimental design or our study
352
00:13:47,180 --> 00:13:50,209
that is our sample means in standard
353
00:13:50,209 --> 00:13:53,570
deviations as well as sample size what
354
00:13:53,570 --> 00:13:55,490
is the obtained level of power that
355
00:13:55,490 --> 00:13:59,660
we've achieved within our study this
356
00:13:59,660 --> 00:14:01,190
should be enough to prepare you to not
357
00:14:01,190 --> 00:14:03,350
only complete the homework given the
358
00:14:03,350 --> 00:14:05,270
effect sizes and other calculations as
359
00:14:05,270 --> 00:14:06,680
well as the values that are provided in
360
00:14:06,680 --> 00:14:09,110
the homework problems themselves but
361
00:14:09,110 --> 00:14:10,550
it's also going to start to familiarize
362
00:14:10,550 --> 00:14:12,110
you with G power so that you can do a
363
00:14:12,110 --> 00:14:13,940
power analysis for your own upcoming
364
00:14:13,940 --> 00:14:17,810
project now again the basic logic is the
365
00:14:17,810 --> 00:14:19,760
same even though exactly the boxes you
366
00:14:19,760 --> 00:14:20,690
click on are going to be slightly
367
00:14:20,690 --> 00:14:22,370
different for your study depending on
368
00:14:22,370 --> 00:14:24,170
the exact nature of the design but
369
00:14:24,170 --> 00:14:25,520
that's something your lab instructor is
370
00:14:25,520 --> 00:14:26,570
going to walk you through as well
371
00:14:26,570 --> 00:14:28,970
depending on whether or not for example
372
00:14:28,970 --> 00:14:32,029
you have a factorial design or situation
373
00:14:32,029 --> 00:14:33,200
we have a single independent variable
374
00:14:33,200 --> 00:14:35,589
with multiple levels so forth and so on
375
00:14:35,589 --> 00:14:37,970
so thanks for tuning in today and good
376
00:14:37,970 --> 00:14:40,660
luck with Alan work
27419
Can't find what you're looking for?
Get subtitles in any language from opensubtitles.com, and translate them here.